leva 0.1.5 → 0.1.7

data/app/views/leva/workbench/new.html.erb

@@ -1,38 +1,81 @@
-<% content_for :title, 'New Experiment' %>
-<div class="px-4 sm:px-6 lg:px-8">
-  <div class="sm:flex sm:items-center">
-    <div class="sm:flex-auto">
-      <h1 class="text-2xl font-semibold text-gray-900">New Experiment</h1>
-      <p class="mt-2 text-sm text-gray-700">Create a new experiment in your workbench.</p>
-    </div>
-  </div>
-  <div class="mt-8 max-w-xl">
-    <%= form_with(model: @experiment, url: experiments_path, local: true, class: "space-y-8 divide-y divide-gray-200") do |form| %>
-      <div class="space-y-8 divide-y divide-gray-200">
-        <div>
-          <div>
-            <h3 class="text-lg leading-6 font-medium text-gray-900">Experiment Information</h3>
-            <p class="mt-1 text-sm text-gray-500">Provide details for your new experiment.</p>
-          </div>
-          <div class="mt-6 grid grid-cols-1 gap-y-6 gap-x-4 sm:grid-cols-6">
-            <div class="sm:col-span-4">
-              <%= form.label :name, class: "block text-sm font-medium text-gray-700" %>
-              <div class="mt-1">
-                <%= form.text_field :name, class: "shadow-sm focus:ring-indigo-500 focus:border-indigo-500 block w-full sm:text-sm border-gray-300 rounded-md" %>
-              </div>
-            </div>
-            <div class="sm:col-span-2">
-              <%= form.label :dataset_id, class: "block text-sm font-medium text-gray-700" %>
-              <div class="mt-1">
-                <%= form.select :dataset_id, options_for_select(@datasets.map { |dataset| [dataset.name, dataset.id] }), class: "shadow-sm focus:ring-indigo-500 focus:border-indigo-500 block w-full sm:text-sm border-gray-300 rounded-md" %>
-              </div>
-            </div>
-          </div>
-        </div>
-      </div>
-      <div class="pt-5">
-        <%= form.submit class: "btn btn-primary btn-block" %>
+<% content_for :title, "New Prompt" %>
+<div class="container mx-auto px-4 py-8 bg-gray-950 text-white">
+  <h1 class="text-3xl font-bold text-indigo-400 mb-6">New Prompt</h1>
+  <%= form_with(model: @prompt, url: workbench_index_path, local: true, class: "bg-gray-800 rounded-lg shadow-lg p-6", data: { controller: "prompt-selector" }) do |form| %>
+    <% if @prompt.errors.any? %>
+      <div class="bg-red-900 border border-red-700 text-red-100 px-4 py-3 rounded-lg mb-4">
+        <h2><%= pluralize(@prompt.errors.count, "error") %> prohibited this prompt from being saved:</h2>
+        <ul>
+          <% @prompt.errors.full_messages.each do |message| %>
+            <li><%= message %></li>
+          <% end %>
+        </ul>
       </div>
     <% end %>
-  </div>
-</div>
+    <div class="mb-4">
+      <%= form.label :name, class: "block text-sm font-semibold mb-2 text-indigo-300" %>
+      <%= form.text_field :name, autofocus: true, class: "w-full bg-gray-700 text-white p-3 rounded-lg focus:ring-2 focus:ring-indigo-500 focus:outline-none" %>
+    </div>
+    <div class="mb-4">
+      <%= form.label :system_prompt, class: "block text-sm font-semibold mb-2 text-indigo-300" %>
+      <%= form.text_area :system_prompt, rows: 2, class: "w-full bg-gray-700 text-white p-3 rounded-lg focus:ring-2 focus:ring-indigo-500 focus:outline-none" %>
+    </div>
+    <div class="mb-4">
+      <%= form.label :predefined_prompt, "Select Predefined Prompt", class: "block text-sm font-semibold mb-2 text-indigo-300" %>
+      <%= form.select :predefined_prompt,
+          options_for_select([['Custom Prompt', '']] + @predefined_prompts.map { |name, content| [name, content] }),
+          {},
+          class: "w-full bg-gray-700 text-white p-3 rounded-lg focus:ring-2 focus:ring-indigo-500 focus:outline-none",
+          data: { action: "change->prompt-selector#toggleUserPrompt" }
+      %>
+    </div>
+    <div class="mb-4" data-prompt-selector-target="userPromptField">
+      <%= form.label :user_prompt, class: "block text-sm font-semibold mb-2 text-indigo-300" %>
+      <%= form.text_area :user_prompt, rows: 5, class: "w-full bg-gray-700 text-white p-3 rounded-lg focus:ring-2 focus:ring-indigo-500 focus:outline-none" %>
+    </div>
+    <div class="mb-4 hidden" data-prompt-selector-target="promptPreview">
+      <h3 class="text-lg font-semibold mb-2 text-indigo-300">Prompt Preview</h3>
+      <div class="bg-gray-700 text-white p-3 rounded-lg" data-prompt-selector-target="previewContent"></div>
+    </div>
+    <div class="flex items-center justify-end space-x-4">
+      <%= link_to "Cancel", workbench_index_path, class: "px-3 py-2 rounded-md text-sm font-medium text-gray-300 hover:bg-gray-800 hover:text-white transition-colors duration-150 ease-in-out" %>
+      <%= form.submit "Create Prompt", class: "px-3 py-2 rounded-md text-sm font-medium bg-indigo-600 text-white shadow-lg hover:bg-indigo-700 transition-colors duration-150 ease-in-out" %>
+    </div>
+  <% end %>
+</div>
+<script>
+  (() => {
+    const application = Stimulus.Application.start()
+
+    application.register("prompt-selector", class extends Stimulus.Controller {
+      static targets = ["userPromptField", "promptPreview", "previewContent"]
+
+      toggleUserPrompt(event) {
+        const selectedContent = event.target.value
+        if (selectedContent) {
+          this.userPromptFieldTarget.style.display = 'none'
+          this.promptPreviewTarget.classList.remove('hidden')
+          this.loadPredefinedPrompt(selectedContent)
+        } else {
+          this.userPromptFieldTarget.style.display = 'block'
+          this.promptPreviewTarget.classList.add('hidden')
+          this.clearUserPrompt()
+        }
+      }
+
+      loadPredefinedPrompt(content) {
+        const userPromptTextarea = this.userPromptFieldTarget.querySelector('textarea')
+        userPromptTextarea.value = content
+        this.previewContentTarget.innerHTML = marked.parse(content)
+      }
+
+      clearUserPrompt() {
+        const userPromptTextarea = this.userPromptFieldTarget.querySelector('textarea')
+        userPromptTextarea.value = ''
+        this.previewContentTarget.innerHTML = ''
+      }
+    })
+  })()
+</script>
+<!-- Include marked.js for Markdown parsing -->
+<script src="https://cdn.jsdelivr.net/npm/marked/marked.min.js"></script>

data/config/routes.rb

@@ -1,12 +1,21 @@
 Leva::Engine.routes.draw do
   root 'workbench#index'
 
-  resources :datasets
-  resources :experiments
+  resources :datasets do
+    resources :dataset_records, path: 'records', only: [:index, :show]
+  end
+  resources :experiments, except: [:destroy] do
+    member do
+      post :rerun
+    end
+    resources :runner_results, only: [:show]
+  end
   resources :prompts
-  resources :workbench, only: [:index, :new, :show] do
-    post 'run', on: :collection
-    post 'run_with_evaluation', on: :collection
-    post 'run_evaluator', on: :collection
+  resources :workbench, only: [:index, :new, :create, :edit, :update] do
+    collection do
+      post 'run'
+      post 'run_all_evals'
+      post 'run_evaluator'
+    end
   end
 end

data/db/migrate/20240813172916_create_leva_datasets.rb

@@ -2,6 +2,7 @@ class CreateLevaDatasets < ActiveRecord::Migration[7.2]
   def change
     create_table :leva_datasets do |t|
       t.string :name
+      t.text :description
 
       t.timestamps
     end

data/db/migrate/20240813173035_create_leva_experiments.rb

@@ -2,6 +2,7 @@ class CreateLevaExperiments < ActiveRecord::Migration[7.2]
   def change
     create_table :leva_experiments do |t|
       t.string :name
+      t.text :description
       t.references :dataset, null: false, foreign_key: { to_table: :leva_datasets }
       t.references :prompt, null: true, foreign_key: { to_table: :leva_prompts }
       t.integer :status

data/db/migrate/20240816201419_create_leva_runner_results.rb

@@ -0,0 +1,11 @@
+class CreateLevaRunnerResults < ActiveRecord::Migration[7.2]
+  def change
+    create_table :leva_runner_results do |t|
+      t.references :experiment, null: false, foreign_key: { to_table: :leva_experiments }
+      t.references :dataset_record, null: false, foreign_key: { to_table: :leva_dataset_records }
+      t.text :prediction
+
+      t.timestamps
+    end
+  end
+end

data/db/migrate/20240816201433_update_leva_evaluation_results.rb

@@ -0,0 +1,8 @@
+class UpdateLevaEvaluationResults < ActiveRecord::Migration[7.2]
+  def change
+    add_reference :leva_evaluation_results, :runner_result, null: false, foreign_key: { to_table: :leva_runner_results }
+    add_column :leva_evaluation_results, :evaluator_class, :string, null: false
+    remove_column :leva_evaluation_results, :prediction, :string
+    remove_column :leva_evaluation_results, :label, :string
+  end
+end

data/db/migrate/20240821163608_make_experiment_optional_for_runner_results.rb

@@ -0,0 +1,6 @@
+class MakeExperimentOptionalForRunnerResults < ActiveRecord::Migration[7.2]
+  def change
+    change_column_null :leva_runner_results, :experiment_id, true
+    change_column_null :leva_evaluation_results, :experiment_id, true
+  end
+end

data/db/migrate/20240821181934_add_prompt_to_leva_runner_results.rb

@@ -0,0 +1,5 @@
+class AddPromptToLevaRunnerResults < ActiveRecord::Migration[7.2]
+  def change
+    add_reference :leva_runner_results, :prompt, null: false, foreign_key: { to_table: :leva_prompts }
+  end
+end

data/db/migrate/20240821183153_add_runner_and_evaluator_to_leva_experiments.rb

@@ -0,0 +1,6 @@
+class AddRunnerAndEvaluatorToLevaExperiments < ActiveRecord::Migration[7.2]
+  def change
+    add_column :leva_experiments, :runner_class, :string
+    add_column :leva_experiments, :evaluator_classes, :text
+  end
+end

data/db/migrate/20240821191713_add_actual_result_to_leva_dataset_records.rb

@@ -0,0 +1,5 @@
+class AddActualResultToLevaDatasetRecords < ActiveRecord::Migration[7.2]
+  def change
+    add_column :leva_runner_results, :actual_result, :text
+  end
+end

data/db/migrate/20240822143201_remove_actual_result_from_leva_runner_results.rb

@@ -0,0 +1,5 @@
+class RemoveActualResultFromLevaRunnerResults < ActiveRecord::Migration[7.2]
+  def change
+    remove_column :leva_runner_results, :actual_result, :text
+  end
+end

data/db/migrate/20240912183556_add_runner_class_to_leva_runner_results.rb

@@ -0,0 +1,5 @@
+class AddRunnerClassToLevaRunnerResults < ActiveRecord::Migration[7.2]
+  def change
+    add_column :leva_runner_results, :runner_class, :string
+  end
+end

data/lib/generators/leva/templates/eval.rb.erb

@@ -1,15 +1,14 @@
 # frozen_string_literal: true
 
 class <%= class_name %>Eval < Leva::BaseEval
-  # @param prediction [String] The prediction to evaluate
-  # @param record [YourRecordClass] The record to evaluate
-  # @return [Leva::Result] The result of the evaluation
-  def evaluate(prediction, record)
+  # @param runner_result [Leva::RunnerResult] The runner result to evaluate
+  # @param recordable [YourRecordClass] The recordable object to evaluate
+  # @return [Float] The score of the evaluation
+  def evaluate(runner_result, recordable)
     # Implement your evaluation logic here
+    # You can access the ground truth using recordable.ground_truth
 
-    Leva::Result.new(
-      label: "<%= file_name.underscore %>",
-      score: score
-    )
+    # Example implementation:
+    runner_result.parsed_predictions.first == recordable.ground_truth ? 1.0 : 0.0
   end
 end

data/lib/generators/leva/templates/runner.rb.erb

@@ -8,4 +8,29 @@ class <%= class_name %>Run < Leva::BaseRun
     # This could involve calling an API, running a local model, etc.
     # Return the result of the run to be used to evaluate the model
   end
+
+  # Uncomment and modify this method to customize parsed predictions
+  # @param runner_result [Leva::RunnerResult] The runner result to parse
+  # @return [Array<String>] The parsed predictions
+  # def parsed_predictions(runner_result)
+  #   # Example: Extract predictions from XML-like tags
+  #   runner_result.prediction.scan(/<prediction>(.*?)<\/prediction>/).flatten
+  # end
+
+  # Uncomment and modify this method to customize ground truth extraction
+  # @param runner_result [Leva::RunnerResult] The runner result to get ground truth from
+  # @return [String] The ground truth for the runner result
+  # def ground_truth(runner_result)
+  #   # Example: Extract ground truth from a specific field
+  #   runner_result.dataset_record.recordable.custom_ground_truth_field
+  # end
+
+  # Uncomment and modify this method to customize regex extraction
+  # @param runner_result [Leva::RunnerResult] The runner result to extract regex from
+  # @return [Regexp, nil] The regex pattern to use for parsing predictions
+  # def extract_regex_pattern(runner_result)
+  #   # Your custom regex extraction logic here
+  #   # For example:
+  #   # /\<result\>(.*?)\<\/result\>/
+  # end
 end

data/lib/leva/version.rb

@@ -1,3 +1,3 @@
 module Leva
-  VERSION = "0.1.5"
+  VERSION = "0.1.7"
 end

data/lib/leva.rb

@@ -1,5 +1,6 @@
 require "leva/version"
 require "leva/engine"
+require "liquid"
 
 module Leva
   # Runs an evaluation experiment with the given run and evals.
@@ -9,9 +10,34 @@ module Leva
   # @param evals [Array<Leva::BaseEval>] The evaluation implementations to use.
   # @return [void]
   def self.run_evaluation(experiment:, run:, evals:)
-    results = run.run(experiment)
+    experiment.update(status: :running)
+
+    experiment.dataset.dataset_records.find_each do |dataset_record|
+      runner_result = run.execute_and_store(experiment, dataset_record, experiment.prompt)
+
+      evals.each do |eval|
+        eval.evaluate_and_store(experiment, runner_result)
+      end
+    end
+
+    experiment.update(status: :completed)
+  rescue StandardError => e
+    experiment.update(status: :failed)
+    Rails.logger.error "Error in experiment #{experiment.name}: #{e.message}"
+  end
+
+  # Runs a single evaluation for a dataset record
+  #
+  # @param experiment [Leva::Experiment] The experiment to run.
+  # @param run [Leva::BaseRun] The run implementation to use.
+  # @param evals [Array<Leva::BaseEval>] The evaluation implementations to use.
+  # @param dataset_record [Leva::DatasetRecord] The dataset record to process.
+  # @return [void]
+  def self.run_single_evaluation(experiment:, run:, evals:, dataset_record:)
+    runner_result = run.execute_and_store(experiment, dataset_record, experiment.prompt)
+
     evals.each do |eval|
-      eval.evaluate_all(experiment, results)
+      eval.evaluate_and_store(experiment, runner_result)
     end
   end
 
@@ -29,20 +55,47 @@ module Leva
       raise NotImplementedError, "#{self.class} must implement #execute"
     end
 
-    # Runs the model on all records in an experiment.
+    # Executes the run on a given dataset record and stores the result.
     #
-    # @param experiment [Leva::Experiment] The experiment to run.
-    # @return [Hash] A hash mapping dataset_record_ids to their execution results.
-    def run(experiment)
+    # @param experiment [Leva::Experiment, nil] The experiment being run, if any.
+    # @param dataset_record [Leva::DatasetRecord] The dataset record to run the model on.
+    # @param prompt [Leva::Prompt] The prompt to store the version of.
+    # @return [Leva::RunnerResult] The stored runner result.
+    def execute_and_store(experiment, dataset_record, prompt)
+      # Expose these to the subclass execution
       @experiment = experiment
-      @prompt = experiment.prompt
+      @prompt = prompt
+
+      result = execute(dataset_record.recordable)
+      RunnerResult.create!(
+        experiment: experiment,
+        dataset_record: dataset_record,
+        prompt: prompt,
+        prediction: result,
+        runner_class: self.class.name
+      )
+    end
 
-      results = {}
-      experiment.dataset.dataset_records.find_each do |dataset_record|
-        result = execute(dataset_record.recordable)
-        results[dataset_record.id] = result
+    # @param runner_result [Leva::RunnerResult] The runner result to parse
+    # @return [Array<String>] The parsed predictions
+    def parsed_predictions(runner_result)
+      if extract_regex_pattern(runner_result)
+        runner_result.prediction.scan(extract_regex_pattern(runner_result)).map { |match| match.first&.strip }.compact
+      else
+        [runner_result.prediction]
       end
-      results
+    end
+
+    # @param runner_result [Leva::RunnerResult] The runner result to extract regex from
+    # @return [Regexp, nil] The regex pattern to use for parsing predictions
+    def extract_regex_pattern(runner_result)
+      runner_result.dataset_record.recordable.extract_regex_pattern if runner_result.dataset_record.recordable.respond_to?(:extract_regex_pattern)
+    end
+
+    # @param runner_result [Leva::RunnerResult] The runner result to get ground truth from
+    # @return [String] The ground truth for the runner result
+    def ground_truth(runner_result)
+      runner_result.dataset_record.ground_truth
     end
   end
 
@@ -51,47 +104,34 @@ module Leva
   # @abstract Subclass and override {#evaluate} to implement
   #   custom evaluation logic.
   class BaseEval
-    # Evaluates the model's prediction against the expected result.
+    # Evaluates the model's prediction against the ground truth.
     #
     # @param prediction [Object] The model's prediction.
-    # @param record [Object] The expected result.
-    # @return [Leva::Result] The evaluation result.
+    # @param recordable [Object] The recordable object containing the ground truth.
+    # @return [Float] The evaluation score.
     # @raise [NotImplementedError] if the method is not implemented in a subclass.
-    def evaluate(prediction, record)
+    def evaluate(prediction, recordable)
       raise NotImplementedError, "#{self.class} must implement #evaluate"
     end
 
-    # Evaluates all results for an experiment.
+    # Evaluates a single runner result and stores the evaluation.
     #
-    # @param experiment [Leva::Experiment] The experiment to evaluate.
-    # @param results [Hash] A hash mapping dataset_record_ids to their execution results.
-    # @return [void]
-    def evaluate_all(experiment, results)
-      experiment.dataset.dataset_records.find_each do |dataset_record|
-        prediction = results[dataset_record.id]
-        evaluation = evaluate(prediction, dataset_record.recordable)
-
-        Leva::EvaluationResult.create!(
-          experiment: experiment,
-          dataset_record: dataset_record,
-          prediction: prediction,
-          score: evaluation.score,
-          label: evaluation.label
-        )
-      end
-    end
-  end
+    # @param experiment [Leva::Experiment, nil] The experiment being evaluated, if any.
+    # @param runner_result [Leva::RunnerResult] The runner result to evaluate.
+    # @return [Leva::EvaluationResult] The stored evaluation result.
+    def evaluate_and_store(experiment, runner_result)
+      @experiment = experiment
+      @runner_result = runner_result
 
-  # Represents the result of an evaluation
-  class Result
-    attr_reader :label, :prediction, :score
+      score = evaluate(runner_result, runner_result.dataset_record.recordable)
 
-    # Initialize a new Result
-    # @param label [String] The label for the result
-    # @param score [Float] The score of the evaluation (0.0 to 1.0)
-    def initialize(label:, score:)
-      @label = label
-      @score = score
+      EvaluationResult.create!(
+        experiment: experiment,
+        dataset_record: runner_result.dataset_record,
+        runner_result: runner_result,
+        score: score,
+        evaluator_class: self.class.name
+      )
     end
   end
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: leva
 version: !ruby/object:Gem::Version
-  version: 0.1.5
+  version: 0.1.7
 platform: ruby
 authors:
 - Kieran Klaassen
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2024-08-14 00:00:00.000000000 Z
+date: 2024-09-12 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: rails
@@ -24,6 +24,20 @@ dependencies:
     - - ">="
       - !ruby/object:Gem::Version
         version: 7.2.0
+- !ruby/object:Gem::Dependency
+  name: liquid
+  requirement: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: 5.5.0
+  type: :runtime
+  prerelease: false
+  version_requirements: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: 5.5.0
 description: Leva is a Ruby on Rails framework for evaluating Language Models (LLMs)
   using ActiveRecord datasets. It provides a flexible structure for creating experiments,
   managing datasets, and implementing various evaluation logic.
@@ -39,26 +53,49 @@ files:
 - app/assets/config/leva_manifest.js
 - app/assets/stylesheets/leva/application.css
 - app/controllers/leva/application_controller.rb
+- app/controllers/leva/dataset_records_controller.rb
 - app/controllers/leva/datasets_controller.rb
 - app/controllers/leva/experiments_controller.rb
+- app/controllers/leva/runner_results_controller.rb
 - app/controllers/leva/workbench_controller.rb
-- app/evals/test_sentiment_accuracy_eval.rb
 - app/helpers/leva/application_helper.rb
+- app/javascript/controllers/prompt_form_controller.js
+- app/javascript/controllers/prompt_selector_controller.js
 - app/jobs/leva/application_job.rb
 - app/jobs/leva/experiment_job.rb
+- app/jobs/leva/run_eval_job.rb
 - app/mailers/leva/application_mailer.rb
+- app/models/concerns/leva/recordable.rb
 - app/models/leva/application_record.rb
 - app/models/leva/dataset.rb
 - app/models/leva/dataset_record.rb
 - app/models/leva/evaluation_result.rb
 - app/models/leva/experiment.rb
 - app/models/leva/prompt.rb
-- app/runners/test_sentiment_run.rb
+- app/models/leva/runner_result.rb
 - app/views/layouts/leva/application.html.erb
+- app/views/leva/dataset_records/index.html.erb
+- app/views/leva/dataset_records/show.html.erb
+- app/views/leva/datasets/_dataset.html.erb
+- app/views/leva/datasets/_form.html.erb
+- app/views/leva/datasets/edit.html.erb
 - app/views/leva/datasets/index.html.erb
+- app/views/leva/datasets/new.html.erb
 - app/views/leva/datasets/show.html.erb
+- app/views/leva/experiments/_experiment.html.erb
+- app/views/leva/experiments/_form.html.erb
+- app/views/leva/experiments/edit.html.erb
 - app/views/leva/experiments/index.html.erb
+- app/views/leva/experiments/new.html.erb
 - app/views/leva/experiments/show.html.erb
+- app/views/leva/runner_results/show.html.erb
+- app/views/leva/workbench/_evaluation_area.html.erb
+- app/views/leva/workbench/_prompt_content.html.erb
+- app/views/leva/workbench/_prompt_form.html.erb
+- app/views/leva/workbench/_prompt_sidebar.html.erb
+- app/views/leva/workbench/_results_section.html.erb
+- app/views/leva/workbench/_top_bar.html.erb
+- app/views/leva/workbench/edit.html.erb
 - app/views/leva/workbench/index.html.erb
 - app/views/leva/workbench/new.html.erb
 - config/routes.rb
@@ -67,12 +104,19 @@ files:
 - db/migrate/20240813173034_create_leva_prompts.rb
 - db/migrate/20240813173035_create_leva_experiments.rb
 - db/migrate/20240813173050_create_leva_evaluation_results.rb
+- db/migrate/20240816201419_create_leva_runner_results.rb
+- db/migrate/20240816201433_update_leva_evaluation_results.rb
+- db/migrate/20240821163608_make_experiment_optional_for_runner_results.rb
+- db/migrate/20240821181934_add_prompt_to_leva_runner_results.rb
+- db/migrate/20240821183153_add_runner_and_evaluator_to_leva_experiments.rb
+- db/migrate/20240821191713_add_actual_result_to_leva_dataset_records.rb
+- db/migrate/20240822143201_remove_actual_result_from_leva_runner_results.rb
+- db/migrate/20240912183556_add_runner_class_to_leva_runner_results.rb
 - lib/generators/leva/eval_generator.rb
 - lib/generators/leva/runner_generator.rb
 - lib/generators/leva/templates/eval.rb.erb
 - lib/generators/leva/templates/runner.rb.erb
 - lib/leva.rb
-- lib/leva/base_eval.rb
 - lib/leva/engine.rb
 - lib/leva/version.rb
 - lib/tasks/auto_annotate_models.rake

data/app/evals/test_sentiment_accuracy_eval.rb

@@ -1,6 +0,0 @@
-class TestSentimentAccuracyEval < Leva::BaseEval
-  def evaluate(prediction, expected)
-    score = prediction == expected ? 1.0 : 0.0
-    Leva::Result.new(label: 'sentiment_accuracy', score: score)
-  end
-end

data/app/runners/test_sentiment_run.rb

@@ -1,13 +0,0 @@
-class TestSentimentRun < Leva::BaseRun
-  def execute(record)
-    # Simple sentiment analysis logic for testing
-    case record.content.downcase
-    when /love|great|excellent/
-      "Positive"
-    when /terrible|bad|awful/
-      "Negative"
-    else
-      "Neutral"
-    end
-  end
-end